1 /* $OpenBSD: diffreg.c,v 1.93 2019/06/28 13:35:00 deraadt Exp $ */
4 * SPDX-License-Identifier: BSD-4-Clause
6 * Copyright (C) Caldera International Inc. 2001-2002.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code and documentation must retain the above
13 * copyright notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed or owned by Caldera
21 * 4. Neither the name of Caldera International, Inc. nor the names of other
22 * contributors may be used to endorse or promote products derived from
23 * this software without specific prior written permission.
25 * USE OF THE SOFTWARE PROVIDED FOR UNDER THIS LICENSE BY CALDERA
26 * INTERNATIONAL, INC. AND CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR
27 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
28 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
29 * IN NO EVENT SHALL CALDERA INTERNATIONAL, INC. BE LIABLE FOR ANY DIRECT,
30 * INDIRECT INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
31 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
32 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
34 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
35 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 * POSSIBILITY OF SUCH DAMAGE.
39 * Copyright (c) 1991, 1993
40 * The Regents of the University of California. All rights reserved.
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)diffreg.c 8.1 (Berkeley) 6/6/93
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
72 #include <sys/capsicum.h>
75 #include <capsicum_helpers.h>
94 * diff - compare two files.
98 * Uses an algorithm due to Harold Stone, which finds
99 * a pair of longest identical subsequences in the two
102 * The major goal is to generate the match vector J.
103 * J[i] is the index of the line in file1 corresponding
104 * to line i file0. J[i] = 0 if there is no
105 * such line in file1.
107 * Lines are hashed so as to work in core. All potential
108 * matches are located by sorting the lines of each file
109 * on the hash (called ``value''). In particular, this
110 * collects the equivalence classes in file1 together.
111 * Subroutine equiv replaces the value of each line in
112 * file0 by the index of the first element of its
113 * matching equivalence in (the reordered) file1.
114 * To save space equiv squeezes file1 into a single
115 * array member in which the equivalence classes
116 * are simply concatenated, except that their first
117 * members are flagged by changing sign.
119 * Next the indices that point into member are unsorted into
120 * array class according to the original order of file0.
122 * The cleverness lies in routine stone. This marches
123 * through the lines of file0, developing a vector klist
124 * of "k-candidates". At step i a k-candidate is a matched
125 * pair of lines x,y (x in file0 y in file1) such that
126 * there is a common subsequence of length k
127 * between the first i lines of file0 and the first y
128 * lines of file1, but there is no such subsequence for
129 * any smaller y. x is the earliest possible mate to y
130 * that occurs in such a subsequence.
132 * Whenever any of the members of the equivalence class of
133 * lines in file1 matable to a line in file0 has serial number
134 * less than the y of some k-candidate, that k-candidate
135 * with the smallest such y is replaced. The new
136 * k-candidate is chained (via pred) to the current
137 * k-1 candidate so that the actual subsequence can
138 * be recovered. When a member has serial number greater
139 * that the y of all k-candidates, the klist is extended.
140 * At the end, the longest subsequence is pulled out
141 * and placed in the array J by unravel
143 * With J in hand, the matches there recorded are
144 * check'ed against reality to assure that no spurious
145 * matches have crept in due to hashing. If they have,
146 * they are broken, and "jackpot" is recorded--a harmless
147 * matter except that a true match for a spuriously
148 * mated line may now be unnecessarily reported as a change.
150 * Much of the complexity of the program comes simply
151 * from trying to minimize core utilization and
152 * maximize the range of doable problems by dynamically
153 * allocating what is needed and reusing what is not.
154 * The core requirements for problems larger than somewhat
155 * are (in words) 2*length(file0) + length(file1) +
156 * 3*(number of k-candidates installed), typically about
157 * 6n words for files of length n.
172 * The following struct is used to record change information when
173 * doing a "context" or "unified" diff. (see routine "change" to
174 * understand the highly mnemonic field names)
177 int a; /* start line in old file */
178 int b; /* end line in old file */
179 int c; /* start line in new file */
180 int d; /* end line in new file */
183 #define diff_output printf
185 static FILE *opentemp(const char *);
186 static void output(char *, FILE *, char *, FILE *, int);
187 static void check(FILE *, FILE *, int);
188 static void range(int, int, const char *);
189 static void uni_range(int, int);
190 static void dump_context_vec(FILE *, FILE *, int);
191 static void dump_unified_vec(FILE *, FILE *, int);
192 static void prepare(int, FILE *, size_t, int);
193 static void prune(void);
194 static void equiv(struct line *, int, struct line *, int, int *);
195 static void unravel(int);
196 static void unsort(struct line *, int, int *);
197 static void change(char *, FILE *, char *, FILE *, int, int, int, int, int *);
198 static void sort(struct line *, int);
199 static void print_header(const char *, const char *);
200 static void print_space(int, int, int);
201 static bool ignoreline_pattern(char *);
202 static bool ignoreline(char *, bool);
203 static int asciifile(FILE *);
204 static int fetch(long *, int, int, FILE *, int, int, int);
205 static int newcand(int, int, int);
206 static int search(int *, int, int);
207 static int skipline(FILE *);
208 static int isqrt(int);
209 static int stone(int *, int, int *, int *, int);
210 static int readhash(FILE *, int);
211 static int files_differ(FILE *, FILE *, int);
212 static char *match_function(const long *, int, FILE *);
213 static char *preadline(int, size_t, off_t);
215 static int *J; /* will be overlaid on class */
216 static int *class; /* will be overlaid on file[0] */
217 static int *klist; /* will be overlaid on file[0] after class */
218 static int *member; /* will be overlaid on file[1] */
220 static int inifdef; /* whether or not we are in a #ifdef block */
222 static int pref, suff; /* length of prefix and suffix */
224 static int anychange;
225 static int hw, padding; /* half width and padding */
227 static long *ixnew; /* will be overlaid on file[1] */
228 static long *ixold; /* will be overlaid on klist */
229 static struct cand *clist; /* merely a free storage pot for candidates */
230 static int clistlen; /* the length of clist */
231 static struct line *sfile[2]; /* shortened by pruning common prefix/suffix */
232 static int (*chrtran)(int); /* translation table for case-folding */
233 static struct context_vec *context_vec_start;
234 static struct context_vec *context_vec_end;
235 static struct context_vec *context_vec_ptr;
237 #define FUNCTION_CONTEXT_SIZE 55
238 static char lastbuf[FUNCTION_CONTEXT_SIZE];
240 static int lastmatchline;
257 diffreg(char *file1, char *file2, int flags, int capsicum)
261 struct pr *pr = NULL;
262 cap_rights_t rights_ro;
269 context_vec_ptr = context_vec_start - 1;
272 * hw excludes padding and make sure when -t is not used,
273 * the second column always starts from the closest tab stop
275 if (diff_format == D_SIDEBYSIDE) {
277 padding = tabsize - (hw % tabsize);
278 if ((flags & D_EXPANDTABS) != 0 || (padding % tabsize == 0))
282 ((padding == MIN_PAD) ? (padding << 1) : padding) - 1;
286 if (flags & D_IGNORECASE)
290 if (S_ISDIR(stb1.st_mode) != S_ISDIR(stb2.st_mode))
291 return (S_ISDIR(stb1.st_mode) ? D_MISMATCH1 : D_MISMATCH2);
292 if (strcmp(file1, "-") == 0 && strcmp(file2, "-") == 0)
295 if (flags & D_EMPTY1)
296 f1 = fopen(_PATH_DEVNULL, "r");
298 if (!S_ISREG(stb1.st_mode)) {
299 if ((f1 = opentemp(file1)) == NULL ||
300 fstat(fileno(f1), &stb1) == -1) {
306 } else if (strcmp(file1, "-") == 0)
309 f1 = fopen(file1, "r");
318 if (flags & D_EMPTY2)
319 f2 = fopen(_PATH_DEVNULL, "r");
321 if (!S_ISREG(stb2.st_mode)) {
322 if ((f2 = opentemp(file2)) == NULL ||
323 fstat(fileno(f2), &stb2) == -1) {
329 } else if (strcmp(file2, "-") == 0)
332 f2 = fopen(file2, "r");
342 pr = start_pr(file1, file2);
345 cap_rights_init(&rights_ro, CAP_READ, CAP_FSTAT, CAP_SEEK);
346 if (caph_rights_limit(fileno(f1), &rights_ro) < 0)
347 err(2, "unable to limit rights on: %s", file1);
348 if (caph_rights_limit(fileno(f2), &rights_ro) < 0)
349 err(2, "unable to limit rights on: %s", file2);
350 if (fileno(f1) == STDIN_FILENO || fileno(f2) == STDIN_FILENO) {
351 /* stding has already been limited */
352 if (caph_limit_stderr() == -1)
353 err(2, "unable to limit stderr");
354 if (caph_limit_stdout() == -1)
355 err(2, "unable to limit stdout");
356 } else if (caph_limit_stdio() == -1)
357 err(2, "unable to limit stdio");
359 caph_cache_catpages();
361 if (caph_enter() < 0)
362 err(2, "unable to enter capability mode");
365 switch (files_differ(f1, f2, flags)) {
377 if (diff_format == D_BRIEF && ignore_pats == NULL &&
378 (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) == 0)
384 if ((flags & D_FORCEASCII) == 0 &&
385 (!asciifile(f1) || !asciifile(f2))) {
390 prepare(0, f1, stb1.st_size, flags);
391 prepare(1, f2, stb2.st_size, flags);
394 sort(sfile[0], slen[0]);
395 sort(sfile[1], slen[1]);
397 member = (int *)file[1];
398 equiv(sfile[0], slen[0], sfile[1], slen[1], member);
399 member = xreallocarray(member, slen[1] + 2, sizeof(*member));
401 class = (int *)file[0];
402 unsort(sfile[0], slen[0], class);
403 class = xreallocarray(class, slen[0] + 2, sizeof(*class));
405 klist = xcalloc(slen[0] + 2, sizeof(*klist));
408 clist = xcalloc(clistlen, sizeof(*clist));
409 i = stone(class, slen[0], member, klist, flags);
413 J = xreallocarray(J, len[0] + 2, sizeof(*J));
418 ixold = xreallocarray(ixold, len[0] + 2, sizeof(*ixold));
419 ixnew = xreallocarray(ixnew, len[1] + 2, sizeof(*ixnew));
420 check(f1, f2, flags);
421 output(file1, f1, file2, f2, flags);
440 * Check to see if the given files differ.
441 * Returns 0 if they are the same, 1 if different, and -1 on error.
442 * XXX - could use code from cmp(1) [faster]
445 files_differ(FILE *f1, FILE *f2, int flags)
447 char buf1[BUFSIZ], buf2[BUFSIZ];
450 if ((flags & (D_EMPTY1|D_EMPTY2)) || stb1.st_size != stb2.st_size ||
451 (stb1.st_mode & S_IFMT) != (stb2.st_mode & S_IFMT))
454 i = fread(buf1, 1, sizeof(buf1), f1);
455 j = fread(buf2, 1, sizeof(buf2), f2);
456 if ((!i && ferror(f1)) || (!j && ferror(f2)))
462 if (memcmp(buf1, buf2, i) != 0)
468 opentemp(const char *f)
470 char buf[BUFSIZ], tempfile[PATH_MAX];
474 if (strcmp(f, "-") == 0)
476 else if ((ifd = open(f, O_RDONLY, 0644)) == -1)
479 (void)strlcpy(tempfile, _PATH_TMP "/diff.XXXXXXXX", sizeof(tempfile));
481 if ((ofd = mkstemp(tempfile)) == -1) {
486 while ((nread = read(ifd, buf, BUFSIZ)) > 0) {
487 if (write(ofd, buf, nread) != nread) {
494 lseek(ofd, (off_t)0, SEEK_SET);
495 return (fdopen(ofd, "r"));
499 splice(char *dir, char *path)
504 dirlen = strlen(dir);
505 while (dirlen != 0 && dir[dirlen - 1] == '/')
507 if ((tail = strrchr(path, '/')) == NULL)
511 xasprintf(&buf, "%.*s/%s", (int)dirlen, dir, tail);
516 prepare(int i, FILE *fd, size_t filesize, int flags)
524 sz = MIN(filesize, SIZE_MAX) / 25;
528 p = xcalloc(sz + 3, sizeof(*p));
529 for (j = 0; (h = readhash(fd, flags));) {
532 p = xreallocarray(p, sz + 3, sizeof(*p));
545 for (pref = 0; pref < len[0] && pref < len[1] &&
546 file[0][pref + 1].value == file[1][pref + 1].value;
549 for (suff = 0; suff < len[0] - pref && suff < len[1] - pref &&
550 file[0][len[0] - suff].value == file[1][len[1] - suff].value;
553 for (j = 0; j < 2; j++) {
554 sfile[j] = file[j] + pref;
555 slen[j] = len[j] - pref - suff;
556 for (i = 0; i <= slen[j]; i++)
557 sfile[j][i].serial = i;
562 equiv(struct line *a, int n, struct line *b, int m, int *c)
567 while (i <= n && j <= m) {
568 if (a[i].value < b[j].value)
570 else if (a[i].value == b[j].value)
581 while (b[j + 1].value == b[j].value) {
589 /* Code taken from ping.c */
598 do { /* newton was a stinker */
603 } while ((x - y) > 1 || (x - y) < -1);
609 stone(int *a, int n, int *b, int *c, int flags)
612 int oldc, tc, oldl, sq;
613 u_int numtries, bound;
615 if (flags & D_MINIMAL)
619 bound = MAX(256, sq);
623 c[0] = newcand(0, 0, 0);
624 for (i = 1; i <= n; i++) {
633 if (y <= clist[oldc].y)
639 if (clist[c[l]].y <= y)
642 c[l] = newcand(i, y, oldc);
647 c[l] = newcand(i, y, oldc);
651 } while ((y = b[++j]) > 0 && numtries < bound);
657 newcand(int x, int y, int pred)
661 if (clen == clistlen) {
662 clistlen = clistlen * 11 / 10;
663 clist = xreallocarray(clist, clistlen, sizeof(*clist));
673 search(int *c, int k, int y)
677 if (clist[c[k]].y < y) /* quick look for typical case */
702 for (i = 0; i <= len[0]; i++)
703 J[i] = i <= pref ? i :
704 i > len[0] - suff ? i + len[1] - len[0] : 0;
705 for (q = clist + p; q->y != 0; q = clist + q->pred)
706 J[q->x + pref] = q->y + pref;
710 * Check does double duty:
711 * 1. ferret out any fortuitous correspondences due
712 * to confounding by hashing (which result in "jackpot")
713 * 2. collect random access indexes to the two files
716 check(FILE *f1, FILE *f2, int flags)
718 int i, j, jackpot, c, d;
724 ixold[0] = ixnew[0] = 0;
727 for (i = 1; i <= len[0]; i++) {
729 ixold[i] = ctold += skipline(f1);
733 ixnew[j] = ctnew += skipline(f2);
736 if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS|D_IGNORECASE|D_STRIPCR)) {
741 * GNU diff ignores a missing newline
742 * in one file for -b or -w.
744 if (flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) {
745 if (c == EOF && d == '\n') {
748 } else if (c == '\n' && d == EOF) {
755 if (flags & D_STRIPCR && (c == '\r' || d == '\r')) {
757 if ((c = getc(f1)) == '\n') {
764 if ((d = getc(f2)) == '\n') {
772 if ((flags & D_FOLDBLANKS) && isspace(c) &&
778 } while (isspace(c = getc(f1)));
783 } while (isspace(d = getc(f2)));
784 } else if ((flags & D_IGNOREBLANKS)) {
785 while (isspace(c) && c != '\n') {
789 while (isspace(d) && d != '\n') {
794 if (chrtran(c) != chrtran(d)) {
797 if (c != '\n' && c != EOF)
798 ctold += skipline(f1);
799 if (d != '\n' && c != EOF)
800 ctnew += skipline(f2);
803 if (c == '\n' || c == EOF)
810 if ((c = getc(f1)) != (d = getc(f2))) {
813 if (c != '\n' && c != EOF)
814 ctold += skipline(f1);
815 if (d != '\n' && c != EOF)
816 ctnew += skipline(f2);
819 if (c == '\n' || c == EOF)
827 for (; j <= len[1]; j++) {
828 ixnew[j] = ctnew += skipline(f2);
832 * fprintf(stderr, "jackpot\n");
836 /* shellsort CACM #201 */
838 sort(struct line *a, int n)
840 struct line *ai, *aim, w;
845 for (j = 1; j <= n; j *= 2)
847 for (m /= 2; m != 0; m /= 2) {
849 for (j = 1; j <= k; j++) {
850 for (ai = &a[j]; ai > a; ai -= m) {
853 break; /* wraparound */
854 if (aim->value > ai[0].value ||
855 (aim->value == ai[0].value &&
856 aim->serial > ai[0].serial))
858 w.value = ai[0].value;
859 ai[0].value = aim->value;
860 aim->value = w.value;
861 w.serial = ai[0].serial;
862 ai[0].serial = aim->serial;
863 aim->serial = w.serial;
870 unsort(struct line *f, int l, int *b)
874 a = xcalloc(l + 1, sizeof(*a));
875 for (i = 1; i <= l; i++)
876 a[f[i].serial] = f[i].value;
877 for (i = 1; i <= l; i++)
887 for (i = 1; (c = getc(f)) != '\n' && c != EOF; i++)
893 output(char *file1, FILE *f1, char *file2, FILE *f2, int flags)
895 int i, j, m, i0, i1, j0, j1, nc;
901 J[m + 1] = len[1] + 1;
902 if (diff_format != D_EDIT) {
903 for (i0 = 1; i0 <= m; i0 = i1 + 1) {
904 while (i0 <= m && J[i0] == J[i0 - 1] + 1){
905 if (diff_format == D_SIDEBYSIDE &&
906 suppress_common != 1) {
907 nc = fetch(ixold, i0, i0, f1, '\0',
910 (hw - nc) + (padding << 1) + 1,
912 fetch(ixnew, J[i0], J[i0], f2, '\0',
920 while (i1 < m && J[i1 + 1] == 0)
926 * When using side-by-side, lines from both of the
927 * files are printed. The algorithm used by diff(1)
928 * identifies the ranges in which two files differ.
929 * See the change() function below.
930 * The for loop below consumes the shorter range,
931 * whereas one of the while loops deals with the
934 if (diff_format == D_SIDEBYSIDE) {
935 for (i=i0, j=j0; i<=i1 && j<=j1; i++, j++)
936 change(file1, f1, file2, f2, i, i,
940 change(file1, f1, file2, f2,
941 i, i, j+1, j, &flags);
946 change(file1, f1, file2, f2,
947 i+1, i, j, j, &flags);
951 change(file1, f1, file2, f2, i0, i1, j0,
955 for (i0 = m; i0 >= 1; i0 = i1 - 1) {
956 while (i0 >= 1 && J[i0] == J[i0 + 1] - 1 && J[i0] != 0)
960 while (i1 > 1 && J[i1 - 1] == 0)
964 change(file1, f1, file2, f2, i1, i0, j1, j0, &flags);
968 change(file1, f1, file2, f2, 1, 0, 1, len[1], &flags);
969 if (diff_format == D_IFDEF || diff_format == D_GFORMAT) {
972 if ((c = getc(f1)) == EOF)
974 diff_output("%c", c);
978 if (anychange != 0) {
979 if (diff_format == D_CONTEXT)
980 dump_context_vec(f1, f2, flags);
981 else if (diff_format == D_UNIFIED)
982 dump_unified_vec(f1, f2, flags);
987 range(int a, int b, const char *separator)
989 diff_output("%d", a > b ? b : a);
991 diff_output("%s%d", separator, b);
995 uni_range(int a, int b)
998 diff_output("%d,%d", a, b - a + 1);
1000 diff_output("%d", b);
1002 diff_output("%d,0", b);
1006 preadline(int fd, size_t rlen, off_t off)
1011 line = xmalloc(rlen + 1);
1012 if ((nr = pread(fd, line, rlen, off)) == -1)
1013 err(2, "preadline");
1014 if (nr > 0 && line[nr-1] == '\n')
1021 ignoreline_pattern(char *line)
1025 ret = regexec(&ignore_re, line, 0, NULL, 0);
1027 return (ret == 0); /* if it matched, it should be ignored. */
1031 ignoreline(char *line, bool skip_blanks)
1034 if (ignore_pats != NULL && skip_blanks)
1035 return (ignoreline_pattern(line) || *line == '\0');
1036 if (ignore_pats != NULL)
1037 return (ignoreline_pattern(line));
1039 return (*line == '\0');
1040 /* No ignore criteria specified */
1045 * Indicate that there is a difference between lines a and b of the from file
1046 * to get to lines c to d of the to file. If a is greater then b then there
1047 * are no lines in the from file involved and this means that there were
1048 * lines appended (beginning at b). If c is greater than d then there are
1049 * lines missing from the to file.
1052 change(char *file1, FILE *f1, char *file2, FILE *f2, int a, int b, int c, int d,
1055 static size_t max_context = 64;
1061 skip_blanks = (*pflags & D_SKIPBLANKLINES);
1063 if ((diff_format != D_IFDEF || diff_format == D_GFORMAT) &&
1066 if (ignore_pats != NULL || skip_blanks) {
1069 * All lines in the change, insert, or delete must
1070 * match an ignore pattern for the change to be
1073 if (a <= b) { /* Changes and deletes. */
1074 for (i = a; i <= b; i++) {
1075 line = preadline(fileno(f1),
1076 ixold[i] - ixold[i - 1], ixold[i - 1]);
1077 if (!ignoreline(line, skip_blanks))
1081 if (a > b || c <= d) { /* Changes and inserts. */
1082 for (i = c; i <= d; i++) {
1083 line = preadline(fileno(f2),
1084 ixnew[i] - ixnew[i - 1], ixnew[i - 1]);
1085 if (!ignoreline(line, skip_blanks))
1092 if (*pflags & D_HEADER && diff_format != D_BRIEF) {
1093 diff_output("%s %s %s\n", diffargs, file1, file2);
1094 *pflags &= ~D_HEADER;
1096 if (diff_format == D_CONTEXT || diff_format == D_UNIFIED) {
1098 * Allocate change records as needed.
1100 if (context_vec_ptr == context_vec_end - 1) {
1101 ptrdiff_t offset = context_vec_ptr - context_vec_start;
1103 context_vec_start = xreallocarray(context_vec_start,
1104 max_context, sizeof(*context_vec_start));
1105 context_vec_end = context_vec_start + max_context;
1106 context_vec_ptr = context_vec_start + offset;
1108 if (anychange == 0) {
1110 * Print the context/unidiff header first time through.
1112 print_header(file1, file2);
1114 } else if (a > context_vec_ptr->b + (2 * diff_context) + 1 &&
1115 c > context_vec_ptr->d + (2 * diff_context) + 1) {
1117 * If this change is more than 'diff_context' lines from the
1118 * previous change, dump the record and reset it.
1120 if (diff_format == D_CONTEXT)
1121 dump_context_vec(f1, f2, *pflags);
1123 dump_unified_vec(f1, f2, *pflags);
1126 context_vec_ptr->a = a;
1127 context_vec_ptr->b = b;
1128 context_vec_ptr->c = c;
1129 context_vec_ptr->d = d;
1134 switch (diff_format) {
1140 diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1141 if (diff_format == D_NORMAL)
1146 diff_output("%c", a > b ? 'a' : c > d ? 'd' : 'c');
1152 diff_output("a%d %d\n", b, d - c + 1);
1154 diff_output("d%d %d\n", a, b - a + 1);
1156 /* add changed lines */
1157 diff_output("a%d %d\n", b, d - c + 1);
1161 if (diff_format == D_GFORMAT) {
1163 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1164 nc = ixold[a > b ? b : a - 1] - curpos;
1165 for (i = 0; i < nc; i++)
1166 diff_output("%c", getc(f1));
1167 for (walk = group_format; *walk != '\0'; walk++) {
1172 fetch(ixold, a, b, f1, '<', 1, *pflags);
1175 fetch(ixnew, c, d, f2, '>', 0, *pflags);
1178 diff_output("%%%c", *walk);
1183 diff_output("%c", *walk);
1186 if (diff_format == D_SIDEBYSIDE) {
1188 print_space(0, hw + padding , *pflags);
1190 nc = fetch(ixold, a, b, f1, '\0', 1, *pflags);
1191 print_space(nc, hw - nc + padding, *pflags);
1193 diff_output("%c", (a>b)? '>' : ((c>d)? '<' : '|'));
1194 print_space(hw + padding + 1 , padding, *pflags);
1195 fetch(ixnew, c, d, f2, '\0', 0, *pflags);
1198 if (diff_format == D_NORMAL || diff_format == D_IFDEF) {
1199 fetch(ixold, a, b, f1, '<', 1, *pflags);
1200 if (a <= b && c <= d && diff_format == D_NORMAL)
1201 diff_output("---\n");
1203 if (diff_format != D_GFORMAT && diff_format != D_SIDEBYSIDE)
1204 fetch(ixnew, c, d, f2, diff_format == D_NORMAL ? '>' : '\0', 0, *pflags);
1205 if (edoffset != 0 && diff_format == D_EDIT) {
1207 * A non-zero edoffset value for D_EDIT indicates that the
1208 * last line printed was a bare dot (".") that has been
1209 * escaped as ".." to prevent ed(1) from misinterpreting
1210 * it. We have to add a substitute command to change this
1211 * back and restart where we left off.
1214 diff_output("%ds/.//\n", a + edoffset - 1);
1215 b = a + edoffset - 1;
1220 if ((diff_format == D_EDIT || diff_format == D_REVERSE) && c <= d)
1223 diff_output("#endif /* %s */\n", ifdefname);
1229 fetch(long *f, int a, int b, FILE *lb, int ch, int oldfile, int flags)
1231 int i, j, c, lastc, col, nc, newcol;
1236 * When doing #ifdef's, copy down to current line
1237 * if this is the first file, so that stuff makes it to output.
1239 if ((diff_format == D_IFDEF) && oldfile) {
1240 long curpos = ftell(lb);
1241 /* print through if append (a>b), else to (nb: 0 vs 1 orig) */
1242 nc = f[a > b ? b : a - 1] - curpos;
1243 for (i = 0; i < nc; i++)
1244 diff_output("%c", getc(lb));
1248 if (diff_format == D_IFDEF) {
1250 diff_output("#else /* %s%s */\n",
1251 oldfile == 1 ? "!" : "", ifdefname);
1254 diff_output("#ifndef %s\n", ifdefname);
1256 diff_output("#ifdef %s\n", ifdefname);
1258 inifdef = 1 + oldfile;
1260 for (i = a; i <= b; i++) {
1261 fseek(lb, f[i - 1], SEEK_SET);
1262 nc = (f[i] - f[i - 1]);
1263 if (diff_format == D_SIDEBYSIDE && hw < nc)
1265 if ((diff_format != D_IFDEF && diff_format != D_GFORMAT) &&
1267 diff_output("%c", ch);
1268 if (Tflag && (diff_format == D_NORMAL ||
1269 diff_format == D_CONTEXT ||
1270 diff_format == D_UNIFIED))
1272 else if (diff_format != D_UNIFIED)
1276 for (j = 0, lastc = '\0'; j < nc; j++, lastc = c) {
1277 if ((c = getc(lb)) == EOF) {
1278 if (diff_format == D_EDIT ||
1279 diff_format == D_REVERSE ||
1280 diff_format == D_NREVERSE)
1281 warnx("No newline at end of file");
1283 diff_output("\n\\ No newline at end of "
1288 * when using --side-by-side, col needs to be increased
1289 * in any case to keep the columns aligned
1292 if (flags & D_EXPANDTABS) {
1293 newcol = ((col/tabsize)+1)*tabsize;
1295 if (diff_format == D_SIDEBYSIDE)
1298 } while (++col < newcol && j < nc);
1300 if (diff_format == D_SIDEBYSIDE) {
1301 if ((j + tabsize) > nc) {
1316 if (diff_format == D_EDIT && j == 1 && c == '\n'
1319 * Don't print a bare "." line
1320 * since that will confuse ed(1).
1321 * Print ".." instead and set the,
1322 * global variable edoffset to an
1323 * offset from which to restart.
1324 * The caller must check the value
1328 edoffset = i - a + 1;
1331 /* when side-by-side, do not print a newline */
1332 if (diff_format != D_SIDEBYSIDE || c != '\n') {
1333 diff_output("%c", c);
1343 * Hash function taken from Robert Sedgewick, Algorithms in C, 3d ed., p 578.
1346 readhash(FILE *f, int flags)
1353 if ((flags & (D_FOLDBLANKS|D_IGNOREBLANKS)) == 0) {
1354 if (flags & D_IGNORECASE)
1355 for (i = 0; (t = getc(f)) != '\n'; i++) {
1356 if (flags & D_STRIPCR && t == '\r') {
1367 sum = sum * 127 + chrtran(t);
1370 for (i = 0; (t = getc(f)) != '\n'; i++) {
1371 if (flags & D_STRIPCR && t == '\r') {
1382 sum = sum * 127 + t;
1386 switch (t = getc(f)) {
1395 if (space && (flags & D_IGNOREBLANKS) == 0) {
1399 sum = sum * 127 + chrtran(t);
1413 * There is a remote possibility that we end up with a zero sum.
1414 * Zero is used as an EOF marker, so return 1 instead.
1416 return (sum == 0 ? 1 : sum);
1422 unsigned char buf[BUFSIZ];
1429 cnt = fread(buf, 1, sizeof(buf), f);
1430 return (memchr(buf, '\0', cnt) == NULL);
1433 #define begins_with(s, pre) (strncmp(s, pre, sizeof(pre)-1) == 0)
1436 match_function(const long *f, int pos, FILE *fp)
1438 unsigned char buf[FUNCTION_CONTEXT_SIZE];
1440 int last = lastline;
1441 const char *state = NULL;
1444 while (pos > last) {
1445 fseek(fp, f[pos - 1], SEEK_SET);
1446 nc = f[pos] - f[pos - 1];
1447 if (nc >= sizeof(buf))
1448 nc = sizeof(buf) - 1;
1449 nc = fread(buf, 1, nc, fp);
1452 buf[strcspn(buf, "\n")] = '\0';
1453 if (isalpha(buf[0]) || buf[0] == '_' || buf[0] == '$') {
1454 if (begins_with(buf, "private:")) {
1456 state = " (private)";
1457 } else if (begins_with(buf, "protected:")) {
1459 state = " (protected)";
1460 } else if (begins_with(buf, "public:")) {
1462 state = " (public)";
1464 strlcpy(lastbuf, buf, sizeof lastbuf);
1466 strlcat(lastbuf, state,
1468 lastmatchline = pos;
1475 return lastmatchline > 0 ? lastbuf : NULL;
1478 /* dump accumulated "context" diff changes */
1480 dump_context_vec(FILE *f1, FILE *f2, int flags)
1482 struct context_vec *cvp = context_vec_start;
1483 int lowa, upb, lowc, upd, do_output;
1487 if (context_vec_start > context_vec_ptr)
1490 b = d = 0; /* gcc */
1491 lowa = MAX(1, cvp->a - diff_context);
1492 upb = MIN(len[0], context_vec_ptr->b + diff_context);
1493 lowc = MAX(1, cvp->c - diff_context);
1494 upd = MIN(len[1], context_vec_ptr->d + diff_context);
1496 diff_output("***************");
1497 if ((flags & D_PROTOTYPE)) {
1498 f = match_function(ixold, lowa-1, f1);
1500 diff_output(" %s", f);
1502 diff_output("\n*** ");
1503 range(lowa, upb, ",");
1504 diff_output(" ****\n");
1507 * Output changes to the "old" file. The first loop suppresses
1508 * output if there were no changes to the "old" file (we'll see
1509 * the "old" lines as context in the "new" list).
1512 for (; cvp <= context_vec_ptr; cvp++)
1513 if (cvp->a <= cvp->b) {
1514 cvp = context_vec_start;
1519 while (cvp <= context_vec_ptr) {
1525 if (a <= b && c <= d)
1528 ch = (a <= b) ? 'd' : 'a';
1531 fetch(ixold, lowa, b, f1, ' ', 0, flags);
1533 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1534 fetch(ixold, a, b, f1,
1535 ch == 'c' ? '!' : '-', 0, flags);
1540 fetch(ixold, b + 1, upb, f1, ' ', 0, flags);
1542 /* output changes to the "new" file */
1543 diff_output("--- ");
1544 range(lowc, upd, ",");
1545 diff_output(" ----\n");
1548 for (cvp = context_vec_start; cvp <= context_vec_ptr; cvp++)
1549 if (cvp->c <= cvp->d) {
1550 cvp = context_vec_start;
1555 while (cvp <= context_vec_ptr) {
1561 if (a <= b && c <= d)
1564 ch = (a <= b) ? 'd' : 'a';
1567 fetch(ixnew, lowc, d, f2, ' ', 0, flags);
1569 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1570 fetch(ixnew, c, d, f2,
1571 ch == 'c' ? '!' : '+', 0, flags);
1576 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1578 context_vec_ptr = context_vec_start - 1;
1581 /* dump accumulated "unified" diff changes */
1583 dump_unified_vec(FILE *f1, FILE *f2, int flags)
1585 struct context_vec *cvp = context_vec_start;
1586 int lowa, upb, lowc, upd;
1590 if (context_vec_start > context_vec_ptr)
1593 b = d = 0; /* gcc */
1594 lowa = MAX(1, cvp->a - diff_context);
1595 upb = MIN(len[0], context_vec_ptr->b + diff_context);
1596 lowc = MAX(1, cvp->c - diff_context);
1597 upd = MIN(len[1], context_vec_ptr->d + diff_context);
1599 diff_output("@@ -");
1600 uni_range(lowa, upb);
1602 uni_range(lowc, upd);
1604 if ((flags & D_PROTOTYPE)) {
1605 f = match_function(ixold, lowa-1, f1);
1607 diff_output(" %s", f);
1612 * Output changes in "unified" diff format--the old and new lines
1613 * are printed together.
1615 for (; cvp <= context_vec_ptr; cvp++) {
1622 * c: both new and old changes
1623 * d: only changes in the old file
1624 * a: only changes in the new file
1626 if (a <= b && c <= d)
1629 ch = (a <= b) ? 'd' : 'a';
1633 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1634 fetch(ixold, a, b, f1, '-', 0, flags);
1635 fetch(ixnew, c, d, f2, '+', 0, flags);
1638 fetch(ixold, lowa, a - 1, f1, ' ', 0, flags);
1639 fetch(ixold, a, b, f1, '-', 0, flags);
1642 fetch(ixnew, lowc, c - 1, f2, ' ', 0, flags);
1643 fetch(ixnew, c, d, f2, '+', 0, flags);
1649 fetch(ixnew, d + 1, upd, f2, ' ', 0, flags);
1651 context_vec_ptr = context_vec_start - 1;
1655 print_header(const char *file1, const char *file2)
1657 const char *time_format;
1662 struct tm tm1, tm2, *tm_ptr1, *tm_ptr2;
1663 int nsec1 = stb1.st_mtim.tv_nsec;
1664 int nsec2 = stb2.st_mtim.tv_nsec;
1666 time_format = "%Y-%m-%d %H:%M:%S";
1670 tm_ptr1 = localtime_r(&stb1.st_mtime, &tm1);
1671 tm_ptr2 = localtime_r(&stb2.st_mtime, &tm2);
1672 strftime(buf1, 256, time_format, tm_ptr1);
1673 strftime(buf2, 256, time_format, tm_ptr2);
1675 strftime(end1, 10, "%z", tm_ptr1);
1676 strftime(end2, 10, "%z", tm_ptr2);
1677 sprintf(buf1, "%s.%.9d %s", buf1, nsec1, end1);
1678 sprintf(buf2, "%s.%.9d %s", buf2, nsec2, end2);
1680 if (label[0] != NULL)
1681 diff_output("%s %s\n", diff_format == D_CONTEXT ? "***" : "---",
1684 diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "***" : "---",
1686 if (label[1] != NULL)
1687 diff_output("%s %s\n", diff_format == D_CONTEXT ? "---" : "+++",
1690 diff_output("%s %s\t%s\n", diff_format == D_CONTEXT ? "---" : "+++",
1695 * Prints n number of space characters either by using tab
1696 * or single space characters.
1697 * nc is the preceding number of characters
1700 print_space(int nc, int n, int flags) {
1704 if ((flags & D_EXPANDTABS) == 0) {
1705 /* first tabstop may be closer than tabsize */
1706 i = tabsize - (nc % tabsize);
1707 while (col >= tabsize) {
1713 diff_output("%*s", col, "");